[Paper Review] Making AI Philosophical Again: On Philip E. Agre's Legacy

Jethro Masís|ArXiv.org|Dec 27, 2025

Digital Humanities and Scholarship20 references4 citations

TL;DR

The paper analyzes Philip E. Agre’s legacy at the AI–philosophy interface, advocating a critical technical practice and reflexivity, while acknowledging limitations of implementing phenomenological concepts in AI.

ABSTRACT

This paper examines the intellectual legacy of Philip E. Agre by situating his work at the intersection of artificial intelligence, philosophy, and critical theory. It reconstructs Agre's proposal of a critical technical practice, according to which AI should be understood not merely as an engineering discipline but as a form of mathematized philosophy shaped by historically contingent metaphors, assumptions, and discourses. Drawing on Heideggerian phenomenology, especially the distinction between ready-to-hand and present-at-hand, Agre sought to reform AI by emphasizing interaction, embedding, indexicality, and deictic representation over traditional mentalist and representational models. The paper analyzes Agre's attempt to operationalize these ideas through computational implementations such as the Pengi system, highlighting both the philosophical ambition and the technical limitations of programming phenomenological concepts. While acknowledging Agre's success in exposing the hidden philosophical commitments of AI and enriching its conceptual vocabulary, the paper ultimately argues that his project encounters a fundamental impasse: the open and self-disclosing character of human existence articulated by Heidegger cannot be fully captured or programmed without reducing ontological phenomena to ontic mechanisms. Agre's enduring contribution therefore lies less in offering a viable Heideggerian AI than in compelling technical practice to become reflexive, historically conscious, and openly philosophical.

Motivation & Objective

• Motivate a rethinking of AI beyond engineering into a form of mathematized philosophy shaped by historical metaphors and discourses.
• Explain Agre's critical technical practice and its heuristic for embedding phenomenology into AI systems.
• Illustrate how Agre operationalized ideas via computational implementations like Pengi.
• Assess the philosophical ambitions and technical limits of programming phenomenological concepts into AI.

Proposed method

• Synthesize Agre’s proposal of critical technical practice with Heideggerian phenomenology (ready-to-hand vs present-at-hand).
• Analyze AI design emphasis on interaction, embedding, indexicality, and deictic representation over representational models.
• Discuss computational implementations (e.g., Pengi) as case studies of Agre’s program.
• Evaluate the philosophical and technical trade-offs of mechanizing ontological concepts.

Experimental results

Research questions

• RQ1 How can AI be reframed as a form of mathematized philosophy rather than a purely engineering discipline?
• RQ2 What are the consequences and limits of operationalizing phenomenological concepts in AI systems?
• RQ3 In what ways does Agre expose hidden philosophical commitments within AI practice, and how should this influence technical work?

Key findings

• Agre's legacy reframes AI as a critical technical practice rather than a purely engineering domain.
• The emphasis on interaction, embedding, indexicality, and deictic representation challenges traditional mentalist and representational AI models.
• Computational implementations like Pengi illustrate both the philosophical ambition and the technical limitations of programming phenomenology.
• There is a fundamental impasse: Heidegger’s open, self-disclosive human existence cannot be fully captured or programmed as ontic mechanisms.
• Agre’s enduring contribution is to push technical practice toward reflexivity, historical consciousness, and openness to philosophical inquiry.